Why is game theory hated

Critique of game theory

Explain the origin of cooperation? (Study

on the weaknesses of a formal approach)
Eckhart Arnold
Computer models of evolutionary game theory have in the last 20
or 30 years been widely used to study such phenomena as cooperation
and reciprocal altruism. However, the scientic value of these models
remains often rather doubtful. In this paper I try to demonstrate (by
examining several examples) that these models are often indeed empirically imprecise and theoretically shallow. Furthermore, I try to
answer why these models often fail and, finally, what requirements a
model must meet if it is to be of any explanatory relevance.

Table of Contents
1 Introduction

2 The theory of the evolution of cooperation

2.1 What to explain the theory of the evolution of cooperation

claimed. . . . . . . . . . . . . . . . . . . . . . . . . . .
2.2 The shape of the explanations of the theory of evolution of the

Cooperation . . . . . . . . . . . . . . . . . . . . . . . . . .
2.2.1 Axelrod's theory of the evolution of cooperation. .
2.2.2 Sch
uler u
About cooperation among egoists. . . . . .
2.2.3 Deer hunting game instead of the prisoner's dilemma. . . . . . .
2.2.4 Cooperation and reputation. . . . . . . . . . . . . .
2.3 A more successful type of theory for comparison: the logic of collective action. . . . . . . . . . . . . . . . . .


. 5
. 5
. 8
. 11
. 12
. 13

3 The declar
aration deficits of the theory of the evolution of the Ko
3.1 Live and let live in the First World War. . . . . . . . .

3.2 sticklebacks and cichlids. . . . . . . . . . . . . . . . . . .

3.3 When soft science beats hard science: game theory and classic social contract theory. . . . . . .
4 conclusion


5 Appendix: Source Codes and Sample Simulations

5.1 Axelrod: evolution of cooperation. . . . . . . . . . . . . . . 40
5.2 Schuessler: Cooperation on Anonymous Markets. . . . . . . . 41
5.3 Skyrms: Deer Hunting Game and Social Contract. . . . . . . 45
6 Revision History



Evolutionary game theory has been very popular for about 30 years. Origin
Developed by biologists to describe strategic aspects of evolution, such as frequency-dependent selection, it is now used as a sub-discipline of game theory in a wide variety of specialist areas

(Biology, economics,
Sociology, Political Science). The evolutionary game theory received a considerable boost above all from the
Publications by Maynard-Smith and Robert Axelrod in the early 1980s
Years. Axelrod in particular made a significant contribution to developing evolutionary game theory as a scientific approach through the innovative use of computer simulations and the breadth of its application examples
to popularize the most diverse fields. High hopes
associated with this novel approach. Namely, by not tying evolutionary game theory to strict rational choice assumptions

is through which some lines of thought in conventional game theory often

so u
appearing exceedingly artificial, 1 one could expect evolutionary game theory as a refined and more flexible form of game theory to emerge
Open up areas of application w
urde; including possibly those that
a formally exact theory formation so far u
Remained locked at all
were. The latter seemed to apply all the more than the use of computer technology, which was developing rapidly at the same time, formal modeling
both greatly simplified and their capabilities increased dramatically.
Have these partly quite high expectations come true in the meantime?
ullt? The fact that new approaches always call for criticism is on
nothing unusual [2]. But while it is in the fr
uhen days of
evolutionary game theory was still concerned with overcoming legitimate criticism by pointing out that this is a still young discipline
[24, p. 38], which will be further developed m
us, so can u
over 20 years
later such excuses no longer apply. Rather, it is up to the
Time, n
uchtern to take stock of what the simulation-controlled
used evolutionary
Game theory so far has been able to achieve and what benefits one can get from
this approach is still expected in the future d
surf. To contribute to
In the following, I would like to examine what the
Theory of cooperation, as described by Axelrod and refined and expanded by numerous imitators and successors, in

An example is the argument of R

backward induction, the f
Only the prisoner's dilemma, which is often repeated, predicts the collapse of cooperation. The argument of the R
Backward induction applies only if there is unconditional mutual rationality. An evolutionary system converges to this only extremely slowly
Solution, and also the experimental findings are not in agreement with the argument.

Application to different problems. With regard to these application examples, which are predominantly in the historical-political area
are taken, my balance is rather very skeptical:
1. Even evolutionary game theory does not succeed in spontaneously generating mathematically exact sociological theory formation. Formal models will only have a marginalized existence in the social sciences in the future as (albeit indispensable) auxiliary sciences such as social statistics or embedded in
a more comprehensive and essentially non-formal explanatory context.
2. The computer models of evolutionary game theory turn out to be
to a high degree as being explanatory irrelevant [2] by applying them

requires a very precise description of the situation from which the

Explanation of the situation often results without a model.
3. A description of the problem situation is only given in exceptional cases
possible, which is so precise that it allows a formal model u
ur that (1) all
to start at all. Is required here
causally relevant influencing variables are recorded in the model, and that (2)
the corresponding parameters can be measured accurately enough,
to allow stable predictions by the model. If, as is unfortunately often the case, these conditions are not met, then
the explanatory potential of formal models shrinks to the epistemological level of mere metaphors.
4. The mathematical-technical refinement of the formal modeling
and the testing of new versions of computer simulations alone f
clocks neither to an expansion of the scope
the theory of the evolution of cooperation, nor is it supported by ei
ne increase in the explanatory power of the theory rewarded. Impulses f
ur the
Further development of this approach are if u
Only from that at all
Expect empiricism.
Since the examples on which this finding is based
Uses largely on the
Theory of the Evolution of Cooperation, Ber
watch nat

only a part of evolutionary game theory, even if to bef

fear is that the basic problems of evolutionary game theory as a whole

concern, especially if they are outside of biology and economics
is applied. In a way, the weaknesses of evolutionary game theory only reflect the difficulties that attempts of the
exact theory formation in certain areas of science again and again
to encounter.

Now one could object that with the above

Expectations are placed on evolutionary game theory, at the most from
Overly confident representatives of the discipline are nourished while
in reality a much more level-headed attitude prevails. It seems
the evolutionary game theory but to get into the dilemma that it is
can either set ambitious goals that extend up to the claim to discuss problems of social contract theory in game theory
want, which the game theory models do not even begin to do justice, or modestly adopt the point of view of
can withdraw
that models are only models that do not allow any direct conclusions to be drawn about reality, which then inevitably raises the question of
what researching models is really good for.
To be fair, however, it must be admitted that even in areas of science, which as a rule do not allow the formation of exact theories, evolutionary game theory also has a lot to offer on the plus side:
1. The (evolutionary) game theory provides a rich reservoir
Metaphors and model examples that affect our scientific imagination and fantasy f
Enrich only the possible considerably.

2. In particular, the theory of the evolution of cooperation allows

simplified and misconceptions u

About the nature and the result

evolutionary processes, such as survival

the stronger, the Un
possibility of selfish cooperation etc., limited
undet to
to reject.
3. The evolution
are game theory makes a constructive contribution to
Criticism of too narrow rational choice assumptions. It sustains by being egg
ne believ
earthy alternative for a better understanding of it
why in many situations there is no egotistical-rational
Behavior prevails in the long term.

The theory of the evolution of cooperati


If in the following from the theory of the evolution of cooperation the

When we are talking about it, it is mainly Robert Axelrod's theory of evolution

the cooperation meant, as well as their further development and modification

by other authors such as Rudolf Sch
uler, Bryan Skyrms and
other. Now to show what this theory can do and, more importantly
what it cannot do, the theory itself must first be presented
become. It is clear that:

1. What does the theory of the evolution of cooperation want to explain?

2. How does it explain what it claims to explain?

3. Which alternative or competition theories are there?


As for the theory of the evolution of cooperation

to explain
aren claimed

The question of what to explain the theory of the evolution of cooperation
is not easy to answer, as the various authors who have contributed to this theory, 2 with different (and before
all also quite differently modest) explanatory claims
uche occur. In summary, the following three can be roughly summarized
Identify objectives:
1. The theory of the evolution of cooperation should be a general cl

ur providing why u
Cooperative behavior in general at all
traditional as could arise in the cultural3 evolution.
2. At the same time it claims individual declarations f
Just to deliver the most varied of situations in which cooperative behavior is observed
can be.
3. Specifically with regard to political philosophy and especially the
Social contract theory it is a concern of the theory of Evo
lution of the cooperation to show the extent to which cooperation is also under
selfish individuals is conceivable without them being through a central
Violence must be enforced.
All three aspects of the theory of evolution will be of cooperation

in the context of the below

examples come up.
First is to discuss how the theory of the evolution of cooperation

proceeds to explain cooperation.


The following Ausf

clocks st
mainly make use of the representations of Robert
Axelrod [5], Rudolf Sch
uler [24], Brian Skyrms [25, 27] and Ken Binmore [8, 9].
ur the concept of cultural evolution, which is in a Darwinian sense, however
without a strict analogy to the nat
To postulate traditional, i.e. genetic evolution, see: [23]


The shape of the declar

arguments of the theory of evo
lution of cooperation

The theory of the evolution of cooperation exists, as already noted,

in many variants. Most of them have in common that they are
in one way or another to the repeated Prisoner's Dilemma as
Basic situation st
use. In the following I will first briefly describe
like Axelrod the iterated prisoner's dilemma on a theory of evolution

the cooperation has expanded [5], to then also only briefly on some
To enter into variants that seem significant to me.

Axelrod's theory of the evolution of cooperation

In a computer simulation, Axelrod initially had a number of different strategies, which he had received after a public appeal from different authors, compete against each other in the paired prisoner's dilemma, each strategy against each other. In each
Duel was f
Only played through the Prisoner's Dilemma a certain number of rounds (but not known to the strategies) in such a way that
the players in each round had the choice of cooperating or defective
ren, with the strategies following the course of the game so far in this decision
could consider. The payouts that each strategy makes in each round
received were totaled. The winner was the strategy that ended up with the
had the highest average score. (Not about the one that does the most
Opponents.) In two consecutive tournaments of this type,
the Axelrod run
every time the strategy won tit for tat, out of which
Axelrod not entirely wrongly on a special ability of this
Strategy closed [5, pp. 25ff.]. But Axelrod did not leave it at a tournament
in which every strategy competes against every other. In a second step
he extended his computer simulation to an evolutionary one, or more precisely
said, population dynamics simulation.4 He also assumed that successful strategies spread over the long term and less successful ones
Supersede strategies m
owed. The result of a population dynamic
Simulations must now by no means correspond to the tournament result, because such strategies, whose relative success is mainly based on the exploitation of good
Good strategies quickly lose ground in the population dynamic simulation as soon as the good
good strategies have died out
(which they usually do first). Axelrod agreed on one point
the result of the population dynamic simulation with that of the tournament

Simplified examples of this type of population dynamic simulation are given in Appendix 5.1.

however u
cleared: It was also possible in the population dynamic simulation
Tit For Tat prevail [5, p. 43ff.].
This, as done later by other scientists
Similar simulations turned out [8, p. 194ff.], in a certain way accidental, f
Ur Axelrod but still remarkable result, moved him to
to study the properties of this strategy more closely. He presented various
dene considerations
about which properties a strategy is successful

make, most importantly his considerations

for collective stability
at be d
urften [5, p. 50ff.]. A strategy is collectively stable when in a
Population dominated by this strategy, no other strategy
can penetrate. (In doing so, Axelrod wisely avoided the stronger term
the evolution
aren stability
at, because is in the scenario he studied
no strategy actually evolution
ar stable.)

Axelrod tried the results of his computer simulation as well as the additional considerations on empirical examples from various fields of science, biology as well as political science
and apply history. Examples that, in his eyes, theoretically
The examined patterns of cooperation showed, among other things: biological mutualisms [5, p. 80ff.], such as the cooperation of cleaner fish
with their hosts; Formation of coalitions in the sense of mutual purpose b
uncertainties in the
outside of the American Senate [5, p. 5]; the live-and-let-live system that historians are in on some front lines
had documented certain phases of the First World War [5, p. 67ff.].
The patterns of cooperation that emerged in the empirical examples
went partly u
Beyond what his computer simulations revealed. Axelrod saw this less as a weakness in his theory than as one
Opportunity to expand. As some of these examples are given below
are discussed in detail, but will not be discussed further here
On the whole, Axelrod's theory of the evolution of cooperation feeds on

so from three sources:

1. Computer simulations of the repeated paired prisoner's dilemma, which Axelrod interprets quite extensively.

2. Additional considerations,
partly in the form of mathematical proof
clock, but partly also purely pragmatic. For example, when Axelrod
the recommendation f
Only gives up the practice of not strictly playing Tit For Tat, but occasionally foregoing retaliation (in order to break a vicious circle of reciprocal retaliation). This recommendation
is not gestured by his own computer simulations
uses, and

it also results in a strategy that is no longer collectively stable
is. Nevertheless, the recommendation to forego retaliation from time to time is undoubtedly negligible
infrequently, if you put the model aside
letting, introducing real situations in which it is mutual
Cooperation is possible.
3. The consideration of empirical examples, which in part give rise to modifications and extensions of the theory of the evolution of coopera
give tion.
Despite (or perhaps because of) its extraordinary success, Axelrods
Theory of the evolution of cooperation received a lot of criticism. Was criticized
de on the one hand that Axelrod had drawn far-reaching and often hasty conclusions from his computer simulations. In fact laid
later computer simulations under similar, but not the same simulation conditions, in some cases suggest quite different conclusions [9, p. 313ff.].
This revealed a fundamental problem of Axelrod's model, namely its lack of robustness, in that it was already slight
ugly deviations
from the assumed initial situation to qualitatively different results f
clocks. Axelrod also has a lack of Ber
consideration of
Accused of knowledge of classical game theory [9, p. 316]. So results
namely already from the so-called folk theorem that every mus
ter more or less great mutual cooperation in repeated
Playing can be stabilized if you can only handle it appropriately
strong sanction mechanism (in case of doubt persistent irrevocable
Defect from the first deviation from the cooperation pattern)
[9, p. 293ff.]. It is against this background that the success of Tit For Tat appears
only as one possibility among many. This criticism reveals clear weaknesses of Axelrod's computer simulation, but as a criticism of his theory
On the whole, it is not entirely fair to the extent that Axelrod is his preference
of Tit For Tat, as set out above, with additional arguments
had underpinned, which were independent of his game theory simulation. In this respect, Tit For Tat was not just one of many
equivalent solutions to the cooperation problem which the
raising repeated two-person prisoner's dilemma.
On the other hand, some of Axelrods were also judged critically
empirical examples. Because the difficulties of empirical application
but also form the main topic of this lecture, these objections will be discussed later, and then exp
to be discussed every hour. First of all, a
Look at the further development of Axelrod's theory by his successors. (Axelrod himself disregarded his theory

of smaller variants, including a rather interesting one, which is evolution
of strategies with the help of a genetic algorithm is hardly simulated [6]
much further developed.)

ler u
About cooperation among egoists

The variants and extensions of Axelrod's theory are manifold and
numerous. It w
urde too far f
clocks to list them all at this point. 5
Instead, only two variants of the theory of the evolution of the

Cooperation are presented, which represent particularly interesting further developments of the theory.
The first variant is Rudolf Sch
ulers model of cooperation in anonymous markets [24]. Sch
uler uses f
Only his model also des
pairwise repeated prisoner's dilemmas, but it changes the game situation in one crucial point. In Axelrod's model is that
The number of repetitions of the prisoner's dilemma is either fixed
or determined by a probability of termination. This default is after
Axelrod's analysis is vital, because only if the schat
If the future is sufficiently long, cooperative behavior can stabilize in the repeated prisoner's dilemma. Interestingly, it gives way
However, it depends precisely on this requirement. In his model, every player has the option to end the game at any time,
namely, which is even more amazing, without giving him this exit option
direct costs (in the form of a fine or the like) are denied
6 One would think that under these conditions the cooperation
collapses quickly and builds a hit and run strate among players
gie established evolutionarily. But this is precisely not the case, but depending on
Choice of parameters (as always with this type of simulation) k
can be
enforce cooperative strategies even under this condition. Also there
as in Axelrod's simulation, is responsible for the shadow of the future, because Sch
uler's simulation deviates on one more point
from Axelrod's simulation. In Axelrods simulation, each player (in each generation) plays against each other, whereby the frequency of encounters resp.
certain pairings is only determined by the weight of the players in the overall population. Not so with Sch
uler: A player who plays the game
breaks off, f
Only look for a new partner for the next round of the game (within the same generation cycle). That come into question
ur nat
rustic only

just a brief summary [14]. An exec
more accurate bibliography until 1994
can be found at [7]
The program code and the results of such a simulation are given in Appendix 5.2

Players from a pool of free players, i.e. those players who have the
Have also canceled the game after the last round (or their game
interrupted by her partner) [24, p. 66ff.]. Now you can
easy u
consider that this pool of free players after some time
collect uger, because cooperative players will
disproportionately many
once they have found each other, strive for cooperation
to continue as long as possible (until their game is accidentally interrupted by floodplains
becomes, a condition that Sch
uler also built into his simulation
has) so that they remain withdrawn from the pool of free players. Indirectly arise the Betr
In this way there are still costs f
just to use theirs
Exit option, because you are forced (in extreme cases after every round) your
To choose partners from a multitude of strategies, which for the most part vary from one another
ugly and thus made up of strategies that are not very important
let earn.
Overall, the (quite u
surprising and therefore theoretically interesting) finding that cooperation can also arise in anonymous markets without an appeal body. Nat
The occurrence of this phenomenon depends very much on the specific model situation. If you change the
Model situation by changing the control parameters or other factors
considered (it would be conceivable, for example, to expand the simulation by
the possibility of a partner change; in this case d
If no cooperation could be expected without ostracism or reputation), then no longer applies
The result is also different, i.e. it can just as well lead to a complete collapse of the cooperation in anonymous markets. The simulation
thus does not show that something will be the case, only that something will be
could. That throws nat
the question of how substantial the results of
In terms of content, simulations actually are. Because if the result of a simulation is simply to state that something
may or may not be, then it certainly doesn't sound particularly interesting or rich. It is SCH
uler credit that he
remains reasonably aware of this problem [24, p. 91/92]. Different to
Axelrod move to Sch
uler much more self-critical (and thus intellectual
more honest). How does Sch
But then what about his approach?
uler motivates his approach by the Ank
up to a classic
Discussion of sociology, which revolves around the question of whether a society
can be held together, which is based only on the individual egoisms of atomized individuals, as appears to be the consequence of a capitalism that is increasingly gaining dominance in society, which is the
The type of businessman (and thus the rational egoist) seems to be elevated to the level of the authoritative type. While Herbert Spencer still does
ute, if instead of brutal military brave merchants socially

set the tone, bef

Durkheim and Tonnies feared that there would be the necessary social cohesion without clear norms that relate accordingly
strong normative institutions st
use, not allow to be produced [24, p.
9-16]. Sch
uler to be able to contribute
by demonstrating that rational selfishness is cooperative behavior
does not make it logically impossible. It wonders of course
Urlich Sch
uler throws
even this question on [24, p. 91/92], whether the representatives of normativism

such a principle u

owe their reasoning
at all presuppose m
to be able to maintain. You don't have to cooperate among egoists
It is impossible to find the idea that all aspects of social life are regulated by markets unenlightening. in the
In case of doubt, a normativist could always take the stand
withdraw that the circumstances in real society are more likely
lie in the parameter range in which the cooperation in the simulation
On a somewhat deeper level, Sch
uler his approach
through a reference to Thomas Hobbes and the so-called Hobbesian
Problem. This establishes the connection to the social contract philosophy. Sch. Understands the Hobbe's problem
uler doing the
Question whether an orderly coexistence among selfishly thought people is possible without strong central authority [24, p. 44/45]. Thomas Hobbes
held such a central power (and, if possible, one that is not too squeamish
goods) as is well known f
ur necessary. In fairness it must be admitted that
himself Sch
uler very well
What is clear about is that the intention of Hobbesian social contract theory, which is about justification and
Explanation of rule and political order is not limited to solving Hobbes' problem. Hobbesian problem is more likely

a designation f
Only a certain abstractly formulated cooperation problem that is not necessarily closely related to Hobbes' philosophy m
usse [24, p. 146 note 6)]. Other game theorists like
e.g. Bryan Skyrms, express themselves less cautiously and argue that
that the questions that could only be dealt with verbally in Hobbes' time, thanks to evolutionary game theory, meanwhile with modern
scientific methods can be treated. We will care
later on
come back. First of all, it should be noted that Sch
uler provides an interesting extension of the Axelrod model. However, return
the same fundamental methodological problems in his simulation too
again, which can already be found at Axelrod, only that Sch
uler is aware of these issues and discusses them openly.



Deer hunting game instead of the prisoner's dilemma

In his cooperation theory, Axelrod had the prisoner's dilemma more

or less unquestionably as a formal model of cooperation problems
Basically. Even if the prisoner's dilemma is undoubtedly a plausible one
Model, this does not yet answer the question of why one of all things
and only the prisoner's dilemma for analyzing cooperation problems
should use. Indeed, there are other alternatives as well. Brian Skyrms, for example, describes various simulations of the deer hunting game,
which is also suitable to serve as a model for a (albeit less severe) cooperation problem [27]. In the game of deer hunting, in contrast to the prisoner's dilemma, there is mutual cooperation with a higher one
Payout is rewarded as a unilateral defect, but a reciprocal defect
is still better than unilateral cooperation. Accordingly, although mutual cooperation is a Nash equilibrium, it is the strategy of
Defection still has the advantage of being risk-safe, i.e. the players will only cooperate if they can assume that their fellow players
do it too. Are the players in the risk-dominant equation?
weight of the non-cooperation, no single player will switch to the
Can achieve better cooperative equilibrium. Vice versa can
but a single player already through the fragile cooperation equilibrium
destroy one-sided defect.
Assuming a population of players is in the risk-dominant equilibrium of non-cooperation, how can this population then be included in the f
ur all more beneficial cooperative balance switch?
Skyrms uses various models to demonstrate that depending on the
assumed replication mechanism a small number of cooperative
Player can expand spatially in a non-cooperative environment. The
It is not worth discussing details here.7 As with the simulations by Axelrod and Sch
However, the qualitative result depends very much
on the simulation parameters and the model situation. That means
but also that one can only deduce from the abstract model without any direct empirical reference that both are possible, the spread of
Cooperation as well as the enforcement of non-cooperation.
But what does the model do and why should such models and
set up at all? Skyrms justifies his approach as follows:
How do we get from the hunt hare equilibrium to the stag hunt equilibri
around? We could approach the problem in two different ways. We could follow
Hobbes in asking the question in terms of rational self-interest. Or we could

See Appendix 5.3, where the program code and the results of a simulation in the style
are held by Skyrms.


Follow Hume by asking the question in a dynamic setting. We can ask these
questions using modern tools - which are more than Hobbes and Hume had
available, but still less than we need for fully adequate answers. [27, p. 10]

like f
ur Sch
it is easier f
ur Skyrms thus above all to the philosophical principle question of how cooperation is possible, and he refers to
Right to Hobbes and Hume, who as a forerunner the normative or factual
Side of this question have already discussed philosophically. What game theory modeling should do now is to discuss these questions
higher theoretical level. Whether this succeeds will have to be investigated.
First of all, it should be noted that Skyrms is another model variant f
ur the
Discussion of the cooperation problem provides, like Sch
ulers and
Axelrod's model suffers from the difficulty that the initial conditions
relatively willing
We have chosen to make the model very sensitive at the same time
reacts to changes in the initial conditions.

Cooperation and reputation

Of course, there are a number of other variants and modifications of Axelrod's approach. A very important variant relates, for example, to approaches that integrate the reputation factor into the cooperation model [1]. Repuatation enables bonuses f
Only cooperation done
to accumulate and later or against
through other partners back into the game
bring to. This results in the cooperation of the close chronological sequence
of the repetitions of the game and also of the strict commitment to pairs
Decoupled interactions. Figuratively
You could also say that
Reputation the M
unzgeld of cooperation is because the difference between
Cooperation without reputation and cooperation with reputation correspond to a
little that between a barter and a real money economy.
Reputation may also be one of the factors contributing to this
It is only responsible that cooperation to a large extent is typical
is a human phenomenon.
Unfortunately it is not possible to carry out all of these variants here
to be discussed in a timely manner, even if this creates the risk of possible variants
have been passed over to the criticism that later applied to the theory of
Evolution of cooperation ge
is practiced, may not apply, so that too

the overall negative conclusion of this treatise against

about the theory of
Evolution of cooperation would not be justified. (Or, with others

Words: The criticism made here of the theory of the evolution of cooperati
on ge
is practiced, can of course
Urlich everyone shows restricted or even refuted
be, by a variant of the theory, not having the same defects
is afflicted, or in u
convincingly a real empirical problem

lost, of the kind discussed below.)
Even if this is only explained in detail later
is clock, so be here already
once summarily anticipated why the theory of evolution

cooperation is a bad theory:

1. The models on which the theory is based
used are not robust (i.e.
Changes in the parameters within the scope of the measurement accuracy f
qualitatively different results) and therefore not empirical
2. The approach lacks a clear question in the sense of a clear one
Idea of ​​which problems are to be solved with it and how.
(Axelrod, Sch
uler and Skyrms provide the examples f
only this charge.)
3. Model research has opposed
about empirical application
Too strong (especially in real, not just experimental) situations
made independent.
So as not to let the misunderstanding arise that it is
with these objections only about the u
common resentment against mathematical methods in the social sciences (especially those in the
Social statistics are indeed very important), the following is supposed to be a theory
are presented and praised, which are also extensively associated with a species
preoccupied with cooperation problems, which avoids these errors.


A more successful type of theory for comparison: the logic of collective action

Mancur Olson's Logic of Collective Action [21] also deals with ei

ne certain kind of cooperation problems, namely the question of how one
large number of people who have a common interest in the provision of a certain common good can come together and organize to actually make the common good available. In the
Collective nature
It is due to the fact that nobody can be excluded from their use, so that nat
rustic everyone prefers to be
Free riders benefit from the collective good w
urde without even one

To contribute to its provision. This f

then adds to the fact that many
uter not even be provided.
If Olson's logic of collective action is now the theory of Evo

lution of cooperation against

is put over, then the accusation is obvious,

that apple
be compared to pears for examining both theories

despite certain similarities

different questions and

pursue a different approach, so that one follows the logic of the kollekti
ven action certainly not as an alternative to the theory of evolution

who can consider cooperation. At most you could use the Lo

gics of collective action as a theory that is a special case
the evolution of cooperation. With regard to this special case
les (cooperation in the production of collective g
utern) you could then
However, we can speak of competition between the two theories.8 In ours
However, it is less important whether the logic of collective action is a substitute or an alternative f
Just a special case of the
Theory of the evolution of cooperation is. F.
ur us is primarily from

Interest, which is why the logic of collective action is a theory that

since its classic formulation by Olson (who in turn refers to Paul
A. Samuelson's Pure Theory of Public Expenditure st
Uses9) an astonishing
lich success story, which is mainly found in a large number of u
of valuable empirical case studies, while the Axelrods method
Evolution of the cooperation also found numerous imitators

has without them even remotely confirming it to the same extent
had found solid empirical research results.
To clarify this question in more detail, it is worth the logic of the collective

Take a closer look at the action. The logic of the collective Han
delns is based on, similar to the theory of the evolution of cooperation

a few very simple basic ideas. So f

ur olson out
the assumption
Reasonable rationality that a collective good is only then
will be made available when the add
the additional benefit that an individual receives by contributing to the provision of the good is greater than
the costs incurred through participation in the provision f
only arise for the individual. (It is important that the add
additional benefits with the cost f
the individual is compared because that is the overall benefit of the collective
Good f
only one individual its costs u
goes without saying, there
Otherwise it is not a collective one, basically even u
none at all
Act well more w
urde.) This connection can also be in a very
simple formula expressed
uck be: Ai = Vi C where Ai is the relative

Indeed, Russel Hardin interprets the logic of collective action to be a
Depicts the prisoner's dilemma situation [12, pp. 16ff.]. However, this interpretation is by no means f
ur all collective
uterine problems compelling. In many cases the game of deer hunting would be
a more adequate game theory presentation f
ur collective
uterine problems.
The roots of the theory extend naturally
urlich still further to
uck. It should be mentioned
here about Humes ber
famous example of the villagers who fail to
together f
just to provide for the watering of their pastures, although it is in everyone's interest
location [15, p. 590]. And it would be astonishing if you weren't already with the ancient ones
Philosophers could come across a description of this problem, which is the political
Everyday experience practically imposes itself.


The additional benefit of the i-th individual is derived from the additional
Benefit Vi minus the costs C, which are assumed to be constant, 10. So that the collective good must be made available

Ai> 0 apply. Equivalent to

the condition can also be written as Vi> C or
expressed as Vi / C> 1
be ucked. If this condition is not met, then
special circumstances must arise in order for a collective good to be made available. An essential part of the theory of the collective
Action is devoted to these circumstances and prerequisites, which make the provision of collective G
even if the condition Ai> 0
not req
is ullt. Such a possibility consists in the coupling with by-products, which are only accessible to those who take part in the provision
share in a collective good.
Another important question f
ur the theory of the logic of the collective

Action is the question of when Ai> 0 is given and when it is not. This depends
The relevant research depends to a large extent on the type of collective good
has there
Only different typologies of collective g
developed. A generalization that is considered in many, but not all cases
Rule of thumb g
Ultimately, 11 says that small groups are mostly able
to provide a collective good, while large groups usually do so
are incapable. This relationship applies u
Incidentally, regardless of the
Transaction costs or organizational issues affecting the provision of
make it even more difficult to utter in large groups.
Without it being detailed here
could be clocked is the logic
of collective action as already noted a u
extremely successful and
above all, empirically extremely fruitful theory. What characteristics
this theory contributed to its success?
1. The basic assumptions of the theory and, accordingly, the formal models that can be constructed on this basis,
turn out to be u
extremely robust. Such is the qualitative relationship
between the group size and the ability to provide one
The collective good is largely invariant to
via parameter fluctuation 12
2. The theory of the logic of collective action pursues a clear one


In order not to make the representation unnecessarily complicated, only the
assumed the very simplest case
In terms of the theory of science, this is a ceteris paribus law. Since the ceteris paribus conditions can be specified relatively precisely using the typologies mentioned above, this remains largely unproblematic.
However, it should be noted that, unlike Olson, this connection sometimes only suggests f
ur certain types of collective g
utern applies.


Objective and avail

ugt u
Over a clearly defined area of ​​application. Unlike the theory of the evolution of cooperation

can easily be determined whether a problem is within the scope

this theory falls or not.
3. At the same time, the logic of collective action is also abstract

enough to solve a multitude of different collective good problems (from

Environmental protection to questions of trade union organization) under one
to summarize a unified description. Last but not least on it
is based on their explanatory power.
The last point in particular deserves to be emphasized. It
in fact, it seems that the logic of collective action is just that right
meets the middle level of abstraction by providing a unifying explanation f